I’m mad at myself. I remember seeing a bunch of tweets about something called AutoSpill back in 2021, and completely ignored it being the luddite that I am. But while preparing for the 2022 ChUG Workshop, I revisited the question and realized the errors of my ways. AutoSpill is just great, and I should have let it enter my life earlier.
Developed by Carlos P. Roca and his collaborators, AutoSpill is a simpler, more efficient approach to calculate the spectral overlap matrix. Traditionally, the automated way to calculate the spillover matrix has required the input of the researcher. Specifically, the user had to set up the gates on the positive and negative fractions for each control. Most analysis software actually automated that gating, but the user still had to revisit each gates and, in many cases make corrections. As Laura Johnston pointed out in an earlier post, the position of your gate definitely has an impact on the quality of the spillover matrix.
This is where AutoSpill shines! It basically removes the need for the user to set up gates around the positive and negative fractions. Instead, it performs a linear regression analysis on the full array of dots on the control. And once it’s done with the first iteration, it will proceed to run additional rounds of linear regressions in order to reduce the compensation error that may still be present and optimize the spillover coefficients.
The kicker is that this approach is actually much easier to use, too. As you’ll see in the video below, it removes the need to create and adjust gates on the positive and negative fractions of your controls, with the exception of the initial FSCxSSC gate.
The conditions for the quality of your controls remain the same – they must be awesome.